Parallel Algorithms / Architecture Synthesis, AIZU International Symposium on (1997)
Aizu-Wakamatsu, Fukushima, JAPAN
Mar. 17, 1997 to Mar. 21, 1997
John Gustafson , Ames Laboratory, USDOE
The U.S. high-performance computing community still uses the term "Grand Challenge" for a variety of difficult problems in computational science. Though it was primarily a means of communicating computing goals to nonpractitioners, it also serves the useful purpose of letting practitioners focus on defining goals more carefully. For purposes of Grand Challenge computing, it is essential to have precise goals, and a way of measuring progress toward those goals. Many Grand Challenges have neither. A particularly common error is to measure the size of a computing problem with some integer "N" that represents the number of grid points or the number of particles or some other count of a discrete quantity. Another common error is to use measures of hardware activity, such as floating point operations per second, as a valid goal for an application programmer. This paper presents an approach to measuring the progress of physical simulations that shows promise for putting computational efforts on firmer scientific ground.
J. Gustafson, "The Program of Grand Challenge Problems: Expectations and Results," Parallel Algorithms / Architecture Synthesis, AIZU International Symposium on(PAS), Aizu-Wakamatsu, Fukushima, JAPAN, 1997, pp. 2.